I named fb as a counterexample. I think you can definitely succeed without validating or invalidating your idea first, but it's a much bigger risk, and you are building a very specific kind of company. How many facebooks and twitters are there in the world?
IMHO, you invalidate an idea after you implement it and see the user's reactions. Before then, you cannot know if it will succeed or not.